Prior-art CMOS and related image sensors typically have a dominant fixed-pattern noise caused by the random variation of the thresholds of the source-follower readout amplifiers in the pixel sensors, necessitating a dark-pattern subtraction operation to cancel the pattern. The dark-pattern subtraction is typically implemented either as a dark-frame subtraction using a separately-captured full dark frame, or as a dark-row subtraction or correlated double sampling, using a double read of each row into difference circuits. Each of these ways slows down the chip operation, complicates the readout circuits and timing requirements, and potentially adds new sources of mismatch and noise.
What is desired is a pixel sensor circuit and operating method that removes most or all of the fixed-pattern noise due to threshold variations, at the earliest possible stage, preferably in the pixel sensor cell itself, to eliminate the requirement for dark-frame subtraction, or dark-row subtraction, or correlated double sampling, or to at least reduce the noise in images that do not go through a dark-pattern pattern correction operation.